GH2 GH2 effective sensor size in ETC mode?

D

den59

Well-known member
Hi everybody,

In usual 16:9 video mode, GH2 is 19 mm x 10.7mm (21.81 diagonal)
The question is, what size is the part of the sensor used in ETC mode?
I calculated it to be about 7.31 x 4.12 (8.39 diagonal)
but I might be wrong.
Crop factor is 2.6.

So, any suggestions are very much appreciated :)

Cheers,
Denis.
 
Yes. Every linear dimension should be divided by the crop factor. If you go into 720P mode you get an even smaller effective sensor size as the crop factor for ETC in this mode will be 3.9.
 
While we don't know for absolute certain, I believe that you are correct. That size would make the GH2 effectively a 1/2" sensor (actually about 5% bigger than 1/2", but close enough that we can just call it 1/2".) And when in 720p mode, the sensor size is effectively the same as a 1/3" video camera.
 
Barry and oceaninchdeep,

thank you for making it clear to me and
supporting my calculations.
 
DPReview says the sensor is only 18.89mm on the long dimension, and that pixel density is 5.9mp/cm2, not the ~6.8 these numbers suggest.

But that's about the fourth different set of numbers I've seen, so I'm not any more inclined to believe them than you.
 
18.89 could be the truth, at least on this video, which, by the way, i think should be sticky, Panasonic claims their sensor to be 19x10.7 in 16:9 mode.

Lee Saxon said:
DPReview says the sensor is only 18.89mm on the long dimension, and that pixel density is 5.9mp/cm2, not the ~6.8 these numbers suggest.

But that's about the fourth different set of numbers I've seen, so I'm not any more inclined to believe them than you.
Click to expand...
 
Barry_Green said:
While we don't know for absolute certain, I believe that you are correct. That size would make the GH2 effectively a 1/2" sensor (actually about 5% bigger than 1/2", but close enough that we can just call it 1/2".) And when in 720p mode, the sensor size is effectively the same as a 1/3" video camera.
Click to expand...

It is hard to believe that a lot of the point and shoot cameras use the same 1/2" sensor and cram 12 megapixels or more into that sensor and then scale it all the way down to 2 megapixels for video. What I wouldn't give for the 2 megapixel Camera I had 10 years ago except with 1080p HD video capabilities and the latest and greatest software.
 
That's what makes the GH2 great -- that's what it is! :)

Well, okay, it's still not a true video camera, but dang, in ETC mode it's pretty darn good, and considering the lowest-cost 1/2" sensor video camera on the market is the EX1R, it's pretty amazing.

Even funnier to me is that prior to the whole "DSLR Revolution", people thought of 1/2" as if it was a pretty big chip... :thumbsup:
 
mpgxsvcd said:
What I wouldn't give for the 2 megapixel Camera I had 10 years ago except with 1080p HD video capabilities and the latest and greatest software.
Click to expand...
Ideally, you'd probably want a sensor with about 8 megapixels to produce 1080p video. The individual light-sensitive elements in the sensor's Bayer Filter are arrayed in an RGBG pattern and each element detects only one of the primary colors. The camera produces full-color RGB pixels by averaging groups of sensor elements together to interpolate each color component in the pixel. The averaging algorithm effectively cuts both horizontal and vertical resolution in half, though some of the lost resolution is recovered with an in-camera sharpening filter.
 
8 megapixels would be good if you wanted/needed 4:4:4, but for 4:2:0 the requirements are a lot lighter, 3mp should be more than enough for that. Note: not that we should be clamoring for 4:2:0 over 4:4:4, obviously, but there are always tradeoffs involved. Would the higher color sampling really be a better way to improve the overall image aesthetic, as compared to having pixels that are 3x larger? Maybe not...
 
I think the bottom line is that we need a sharp fast zoom lens in the wide to mid range focal lengths to really make the ETC mode work. 1080p @ 60 FPS would be nice as well for slow motion with it.

Anyone think the 12-50mm F2.5-3.3 is going to become a reality? I actually think the 25mm F1.4 would be even better than the zoom lens. Sure it is super close to the 20mm F1.7. However, the 25mm is said to be a lens that can AF quickly which the 20mm cannot.
 
Barry_Green said:
8 megapixels would be good if you wanted/needed 4:4:4, but for 4:2:0 the requirements are a lot lighter, 3mp should be more than enough for that. Note: not that we should be clamoring for 4:2:0 over 4:4:4, obviously, but there are always tradeoffs involved. Would the higher color sampling really be a better way to improve the overall image aesthetic, as compared to having pixels that are 3x larger? Maybe not...
Click to expand...

Would you all rather have the 16 megapixel 1080p chip of the GH2 with two ETC modes or a 4/3s chip with 8 megapixels and 1920x1080 total resolution. Basically would you trade the multiple focal length ETC modes for a single focal length ideal setup?
 
For a video camera? Definitely fewer pixels and closer to native resolution. I'd give up teleconversion in exchange for pristine images. Well, heck, that's pretty much what the AF100 does -- it's got no ETC, but it has (nearly) alias-free imagery from a 12-megapixel chip. I'd rather have as few pixels as possible, as big as possible. You want as many as you need, but no more than that.

Of course, that would never happen on a stills camera or hybrid; you need the higher pixel count for stills. Hence why the ETC mode is a great little compromise.
 
Barry_Green said:
For a video camera? Definitely fewer pixels and closer to native resolution. I'd give up teleconversion in exchange for pristine images. Well, heck, that's pretty much what the AF100 does -- it's got no ETC, but it has (nearly) alias-free imagery from a 12-megapixel chip. I'd rather have as few pixels as possible, as big as possible. You want as many as you need, but no more than that.

Of course, that would never happen on a stills camera or hybrid; you need the higher pixel count for stills. Hence why the ETC mode is a great little compromise.
Click to expand...

I was surprised to hear the AF100 does not have ETC. I know it is supiorior because it is "nearly" alias free. However, I would think just having the option to use ETC would be better than not having it at all.

On a side note. How does the variable frame rate work on the AF100? Do you actually get to select a specific numeric value for the frame rate? Can you select any value from 1 to 30?
 
Barry_Green said:
8 megapixels would be good if you wanted/needed 4:4:4, but for 4:2:0 the requirements are a lot lighter, 3mp should be more than enough for that.
Click to expand...
Actually, an 8 megapixel sensor would be pretty close to 4:2:0 color resolution at 1080p. The Bayer Filter grid is arranged like this:

G R G R G R G R
B G B G B G B G
G R G R G R G R
B G B G B G B G

Note that there are twice as many G pixels as either R or B pixels. 8 megapixels would be a 3840x2080 array, comprised of a 1920x1080 array of Green pixels and two 960x540 arrays of Red and Blue pixels. YCC luma and chroma signals would be calculated as approximately:

Y = 59% G + 30% R + 11% B
Cr = 50% R + 42% G + 8% B
Cb = 50% B + 33% G + 17% R

Since luma is mostly determined by the G component, it should correlate well with the 1920x1080 resolution of the G elements. Likewise, Cr and Cb chroma signals are mostly determined by the R and B components, whose resolution is 960x540. This is effectively the same resolution you get from 4:2:0 color sampling.
 
mpgxsvcd said:
However, I would think just having the option to use ETC would be better than not having it at all.
Click to expand...
ETC = aliasing in full sensor mode. No two ways around it. The AF100 isn't a stills camera, it doesn't alias, and therefore it doesn't have and can't have ETC.

On a side note. How does the variable frame rate work on the AF100? Do you actually get to select a specific numeric value for the frame rate? Can you select any value from 1 to 30?
Click to expand...
There are 20 selectable frame rates, from 12 up to 60.
 
lpowell said:
8 megapixels would be a 3840x2080 array, comprised of a 1920x1080 array of Green pixels and two 960x540 arrays of Red and Blue pixels.
Click to expand...
But the math doesn't add up that way. In a 3840x2160 array, you'd have 3840 green samples horizontally, although interleaved up/down, but within the context of 1080 high you'd have basically a 3840x1080 matrix of green.

You'd have 1920 x 1080 red, and 1920 x 1080 blue.

Let's put it simply: in an 8-megapixel Bayer sensor, you're going to have 4 mpix of green, and 2mpix of red, and 2mpix of blue. 4+2+2 = 8.

That's why I said that for 4:4:4 you might need 8 megapixels, because you would indeed have a 1920x1080 matrix of red and blue.

For 4:2:0, you could get away with a 4-megapixel sensor, which would give you 960x540 red, 960x540 blue, and 1920x1080 green.
 
Barry_Green said:
For 4:2:0, you could get away with a 4-megapixel sensor, which would give you 960x540 red, 960x540 blue, and 1920x1080 green.
Click to expand...
Right, I added up the pixels incorrectly. In an 8-megapixel, 3840x2160 sensor there would be 2160 rows of 1920 G pixels and 1080 rows of 1920 B and 1920 G pixels. However, a 4-megapixel sensor would have to have a non-square pixel aspect ratio, each pixel half as wide as it is tall...

So it seems mpgxsvcd's 2-megapixel sensor would actually be the closest fit after all. That would contain an array of 960x1080 G pixels and two arrays of 960x540 R and B pixels. With a sharpening filter, it should have a resolution pretty close to 1920x1080 in 4:2:0 color depth.
 
Last edited:
And massive, massive pixels... which should, theoretically, blow the doors off the joint in terms of DR and noise performance...
 
You must log in or register to reply here.
Back
Top